QUAK

QUAK是一个针对韩语-英语神经机器翻译的合成质量评估数据集，由高丽大学创建。该数据集包含三个子集：QUAK-M、QUAK-P和QUAK-H，分别利用单语、平行和混合语料库生成，总条数达到6578002条。QUAK通过全自动方式生成，无需人工干预，支持大规模数据扩展和语言扩展。该数据集主要用于训练和评估机器翻译输出的质量，特别是在没有参考句子的情境下预测翻译质量。

QUAK is a synthetic quality assessment dataset for Korean-English neural machine translation, constructed by Korea University. This dataset includes three subsets: QUAK-M, QUAK-P, and QUAK-H, which are generated using monolingual, parallel, and mixed corpora respectively, with a total of 6,578,002 samples. QUAK is generated in a fully automated fashion without any human intervention, supporting large-scale data and language expansion. This dataset is primarily used for training and evaluating the quality of machine translation outputs, especially for predicting translation quality in scenarios where no reference sentences are available.